org.nd4j.linalg.dataset.api.MultiDataSet Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
package org.nd4j.linalg.dataset.api;
import org.nd4j.linalg.api.ndarray.INDArray;
import java.io.*;
import java.util.List;
/**
* MultiDataSet is an interface for representing complex data sets, that have (potentially) multiple inputs and outputs
* For example, some complex neural network architectures may have multiple independent inputs, and multiple independent
* outputs. These inputs and outputs need not even be the same opType of data: for example, images in and sequences out, etc
*/
public interface MultiDataSet extends Serializable {
/**
* Number of arrays of features/input data in the MultiDataSet
*/
int numFeatureArrays();
/**
* Number of arrays of label/output data in the MultiDataSet
*/
int numLabelsArrays();
/**
* Get all of the input features, as an array of INDArrays
*/
INDArray[] getFeatures();
/**
* Get a single feature/input array
*/
INDArray getFeatures(int index);
/**
* Set all of the features arrays for the MultiDataSet
*/
void setFeatures(INDArray[] features);
/**
* Set a single features array (by index) for the MultiDataSet
*/
void setFeatures(int idx, INDArray features);
/**
* Get all of the labels, as an array of INDArrays
*/
INDArray[] getLabels();
/**
* Get a single label/output array
*/
INDArray getLabels(int index);
/**
* Set all of the labels arrays for the MultiDataSet
*/
void setLabels(INDArray[] labels);
/**
* Set a single labels array (by index) for the MultiDataSet
*/
void setLabels(int idx, INDArray labels);
/**
* Whether there are any mask arrays (features or labels) present for this MultiDataSet
*/
boolean hasMaskArrays();
/**
* Get the feature mask arrays. May return null if no feature mask arrays are present; otherwise,
* any entry in the array may be null if no mask array is present for that particular feature
*/
INDArray[] getFeaturesMaskArrays();
/**
* Get the specified feature mask array. Returns null if no feature mask is present for that particular feature/input
*/
INDArray getFeaturesMaskArray(int index);
/**
* Set the feature mask arrays
*/
void setFeaturesMaskArrays(INDArray[] maskArrays);
/**
* Set a single feature mask array by index
*/
void setFeaturesMaskArray(int idx, INDArray maskArray);
/**
* Get the labels mask arrays. May return null if no labels mask arrays are present; otherwise,
* any entry in the array may be null if no mask array is present for that particular label
*/
INDArray[] getLabelsMaskArrays();
/**
* Get the specified label mask array. Returns null if no label mask is present for that particular feature/input
*/
INDArray getLabelsMaskArray(int index);
/**
* Set the labels mask arrays
*/
void setLabelsMaskArray(INDArray[] labels);
/**
* Set a single labels mask array by index
*/
void setLabelsMaskArray(int idx, INDArray labelsMaskArray);
/**
* Save this MultiDataSet to the specified stream. Stream will be closed after saving.
*/
void save(OutputStream to) throws IOException;
/**
* Save this MultiDataSet to the specified file
*/
void save(File to) throws IOException;
/**
* Load the contents of this MultiDataSet from the specified stream. Stream will be closed after loading.
*/
void load(InputStream from) throws IOException;
/**
* Load the contents of this MultiDataSet from the specified file
*/
void load(File from) throws IOException;
/**
* SplitV the MultiDataSet into a list of individual examples.
*
* @return List of MultiDataSets, each with 1 example
*/
List asList();
/**
* Clone the dataset
*
* @return a clone of the dataset
*/
MultiDataSet copy();
/**
* Set the metadata for this MultiDataSet
* By convention: the metadata can be any serializable object, one per example in the MultiDataSet
*
* @param exampleMetaData Example metadata to set
*/
void setExampleMetaData(List exampleMetaData);
/**
* Get the example metadata, or null if no metadata has been set
* Note: this method results in an unchecked cast - care should be taken when using this!
*
* @param metaDataType Class of the metadata (used for opType information)
* @param Type of metadata
* @return List of metadata objects
*/
List getExampleMetaData(Class metaDataType);
/**
* Get the example metadata, or null if no metadata has been set
*
* @return List of metadata instances
* @see {@link #getExampleMetaData(Class)} for convenience method for types
*/
List getExampleMetaData();
/**
* This method returns memory amount occupied by this MultiDataSet.
*
* @return value in bytes
*/
long getMemoryFootprint();
/**
* This method migrates this MultiDataSet into current Workspace (if any)
*/
void migrate();
/**
* This method detaches this MultiDataSet from current Workspace (if any)
*/
void detach();
/**
* @return True if the MultiDataSet is empty (all features/labels arrays are empty)
*/
boolean isEmpty();
/**
* Shuffle the order of the examples in the MultiDataSet. Note that this generally won't make any difference in
* practice unless the MultiDataSet is later split.
*/
void shuffle();
}